October 8–9, 2026 | In person, Durham NC

We’re delighted to announce the dates for our third annual summit, coming in October 2026! Please stay tuned for additional details to be announced soon.

You can learn more about last year’s summit, including the program and impact report here.

Duke AI Health hosted two virtual seminars designed to support the Duke research community as it transitions from PACE to the Protected Research Compute Cluster (PRCC). The first session, PACE to PRCC: Introduction for School of Medicine Researchers, provided an overview of the PRCC’s architecture and expanded capacity. The second session, PACE to PRCC: Managing Your Project Transition, guided users through what to expect as projects begin moving into the new environment. These virtual seminars were led by a multidisciplinary team of experts from Biostatistics and Bioinformatics, Duke Health Technology Solutions (DHTS) and the School of Medicine, including Ricardo Henao, Jay Stotler, Aby Veiga, Danny Williford, John Bradley, and were hosted by AI Health Managing Director, Shelley Rusincovitch. The recorded content, slides and links are available to the Duke community on the Data Science SharePoint.

LINKS

• The Protected Research Compute Cluster: Introduction for School of Medicine Researchers
• PACE to PRCC: Managing Your Project Transition

Benjamin Goldstein, PhD, Duke AI Health’s Director of Data Science, was among the participants at the Department of Medicine’s second Artificial Intelligence and Medicine Symposium on February 3. The event, which revolved around the theme of AI’s transformative impact on medicine and research, included individual presentations and panel discussions, was capped by a resource fair. Dr. Goldstein took part in a panel focused on AI-centered collaborations in health research that was moderated by Vice Dean for Data Science & AI Christopher Lindsell, PhD.

READ MORE

A new case study led by students and faculty from Columbia University and Duke, including Duke School of Nursing professor and AI Health Faculty Council member Michael Cary, PhD, and doctoral student Thomas Merrill explores how nurses increasingly encounter patients who bring AI-generated health information to the bedside. The paper highlights the ethical and clinical tensions that arise when AI tools interpret medical data without clinical context, and it introduces practical communication strategies nurses can use to navigate these situations. The authors propose two structured approaches—Ask-Balance-Clarify-Document for communication with patients and AI-SBAR for communication with clinical teams —to help nurses address AI-generated information safely and ethically.

READ MORE

AI Health faculty affiliate Chuan Hong, PhD, whose American Heart Association–supported research evaluates the PREVENT equations for 10-year atherosclerotic cardiovascular disease risk, mentored graduate student Ziye Tian. Tian received a Travel Award for Junior Investigators to present her work, “Toward a Deeper Understanding of PREVENT for 10 Year Atherosclerotic Cardiovascular Risk: Subgroup Fairness and Predictive Value of Social Determinants of Health,” at the International Stroke Conference 2026. In extensive analyses using data from Truveta electronic health records, Tian assessed model fairness across demographic and social determinants of health subgroups and examined the added predictive value of those social determinants of health.

READ MORE

Duke AI Health Research Scientist Whitney Welsh, PhD, presented a poster, “The Duke AI Health Community of Practice: A Model for Increasing AI Proficiency and Critical Understanding,” at the Conference on Society-Centered AI in Durham this February. Welsh and co-author Shelley Rusincovitch, MMCi, AI Health managing director, described the process of creating a logic model for the Duke AI Health Community of Practice and shared results from last year’s feedback survey, which show that the Community of Practice is achieving many of its short-term goals.

READ MORE

This article was originally published by the Duke Pratt School of Engineering. You can learn more about Dr. Dunn and his work with the DANNCE 3D mapping program, which was supported by Duke Forge, a precursor organization to AI Health, in an article available at the Duke AI Health website.

The subjects in the research videos created by Timothy Dunn, an assistant professor of biomedical engineering at Duke University, aren’t immediately obvious. Rather than show tufts of fur or swishing tails, the animal models––usually mice or rats––are instead depicted by straight lines and colorful dots that move around an otherwise empty screen.

These videos are created through the program DANNCE, short for 3-Dimensional Aligned Neural Network for Computational Ethology, a tool Dunn and his team developed in 2021. Using videos of freely moving rats, the team trained machine-learning algorithms and neural networks to identify and map the precise 3D locations of the body joints on the animals. Researchers could then relate these measurements to data collected from brain recording technologies to examine links between neuronal activity and specific behaviors.

READ MORE

This past December’s Data Science Poster Showcase, co-hosted by Duke AI Health, Duke Pratt School of Engineering, the Duke Center for Computational Thinking, the Duke Center for Computational and Digital Health Innovation, and Duke Health, included a “People’s Choice” award for best poster presentation. For this category, showcase attendees were able to cast votes for their favorite entries. The Vox Populi Award for 2025 was a tie, with awards going to Logan Bailey and Charles Scales for “Enhancing Translation Efficiency: Evaluating AI Assisted Translation for Clinical and Research Documents,” and to Jaden Sacks and colleagues for “Beyond Advance Directives: A Machine Learning Tool to Aid End-of-Life Decisions in Dementia Care.”

Please join us in congratulating our winners!

LEARN MORE

This March, the Duke AI Health Community of Practice hosted the Machine Learning School: Health AI (MLSS-HealthAI), the 13th in the ML School Series since 2017. Machine learning expert Ricardo Henao, PhD, led the three-day program, which featured 10 expert presenters from across Duke. The event brought together 149 participants representing 58 organizations, including 56 students and 22 scholarship recipients (in keeping with its educational mission, the Community of Practice reserved a minimum of 10% of seats for scholarships to support attendance by those who would otherwise be unable to attend).

A major highlight of this year’s event was the hands-on use of the Duke Health AI Studio, which offers a variety of AI platforms for Duke users in a protected environment. With help and expert guidance from Duke Health Technology Solutions (DHTS), participants were able to explore real-world examples, and apply new techniques in an immersive computing environment.

LEARN MORE

In February, a select steering committee representing multiple schools, institutes and disciplines across Duke University published a report on artificial intelligence to the Duke University Provost. The report synthesized perspectives on Duke’s strengths in AI as well as gaps and opportunities, including the potential for improving institutional approaches to developing, applying, and providing ethical oversight for these technologies at the university. The Steering Committee, whose leadership included Duke AI Health Faculty Council Member Ricardo Henao, PhD, as a co-chair, also examined infrastructural needs for AI at Duke and articulated a strategic vision for the future of human-centered AI in research and education.

READ MORE

On March 29, Duke University Executive Vice President for Health Affairs and School of Medicine Dean Mary E. Klotman, MD, announced that Duke Clinical Research Institute Director of Data Science Christopher Lindsell, PhD, has been appointed to serve as Vice Dean for Data Science & AI at the Duke University School of Medicine, effective April 1. In this new role, Dr. Lindsell will lead the strategy, execution, and governance of AI and data science across the School of Medicine while collaborating closely with key partners in the Duke University Health System and the university. His immediate priorities include advancing a unified strategy for AI research, education, and workforce training, as well as building multidisciplinary teams, coordinating platform development and oversight, and accelerating our learning health system to ensure that discovery, clinical decision‑making, and operations at Duke are informed by the highest-quality data possible. READ MORE

Please help us make AI Health’s Community of Practice better! In this brief survey, we are trying to get a sense of what AI Health offerings you may have found useful or valuable over the past year. Have you participated in any of our classes or seminars? Have you attended one of our conferences or summits? Are you enjoying the Friday Roundup? Are there any ways we can improve or serve you better?

Take the survey here: https://duke.qualtrics.com/jfe/form/SV_3sCUufSFTxCiMsu

The responses from last year’s impact survey were extremely helpful in our planning for this year’s offerings.  We are genuinely interested in what you have to say, and what we can do to improve on what we are doing! Check out last year’s results at https://duke.is/2025-ai-health-survey-summary

The linked Qualtrics survey is anonymous and should only take a few minutes to complete. As an added incentive, participants who complete their surveys may choose to enter a raffle for Duke AI Health mugs and stickers.  We very much appreciate your feedback!

Duke AI Health Program Coordinator Tiffany Torres recently earned her SHRM-CP (Society for Human Resource Management–Certified Professional) credential. This nationally recognized certification reflects advanced expertise in people operations, workplace policies, and employee support—capabilities that strengthen internal team performance while enhancing collaboration with external partners and sponsors. Her accomplishment underscores a commitment to supporting effective, well-coordinated teams and the people-centered practices that enable Duke AI Health’s work at the intersection of AI and health. Please join us in congratulating Tiffany on this achievement!

A paper co-authored by Duke AI Health researchers Christine Shen, Christian Péan,MD, and Samuel Berchuck, PhD, received a Student Paper Award from the Bayesian Statistical Science Section of the American Statistical Association. The work introduces a new spatial modeling approach to identify geographic patterns in hospital readmission risk following upper extremity fracture surgery while accounting for competing mortality risk. Using high-resolution maps, the team highlights local areas where patients face elevated risk, creating opportunities for targeted, community-level interventions. This framework demonstrates how advanced statistical methods can translate complex health data into actionable insights for population health and care delivery.

READ MORE

The Duke AI Health Community of Practice is pleased to host the Duke Machine Learning Spring School 2026: Health AI (MLSS-HealthAI), offered in March as an in-person three-day class that will provide perspective and insight on the fundamentals of generative artificial intelligence methods and applications. The curriculum in the MLSS-HealthAI is targeted to individuals interested in learning about generative AI with health applications and will introduce the mathematics and statistics at the foundation of current generative and representation learning models, plus provide context for the methods that have formed the foundations of rapid growth in AI. Additionally, the MLSS-HealthAI will provide hands-on experience during practical sessions that illustrate the use of such methods and applications. LEARN MORE

In June 2025, scientific collaborators from Duke AI Health and the American Heart Association, led by principal investigator and former AI Health Director Michael Pencina, PhD, concluded a five-year study funded by the National Institutes of Health and designed to improve the ability to predict a patient’s risk of stroke. After aggregating and harmonizing patient-level data collected from four NIH-sponsored observational cohort studies, the study investigators evaluated the performance of existing stroke risk prediction models, developed new models with the potential to improve clinical decision-making, and explored ways to mitigate algorithmic bias and improve the fairness of models in clinical use. Key findings from the study have been shared through 18 publications in high-impact academic journals and conference proceedings, and the code used to conduct analyses is publicly available through the study’s dedicated GitHub repository.

LEARN MORE

This fall, the Duke AI Health Community of Practice successfully concluded its third Scientific Writing Workshop for Staff. This four-unit minicourse is designed to offer a wide-ranging introduction to scientific writing and peer-reviewed publication. AI Health Research Scientist Whitney Welsh, PhD, has created an impact report (linked below) that provides a detailed snapshot of how this most recent iteration of the program was received. In December, workshop participants presented the research posters they developed as a key class project at the 2025 Health Data Science Poster Showcase. Congratulations and thanks to all of our workshop participants for making the fall 2025 offering such a success!

READ MORE

This December, Duke AI Health partnered with the Pratt School of Engineering, the Duke Center for Computational Thinking, the Duke Center for Computational and Digital Health Innovation, and Duke Health to present the fourth Data Science Poster Showcase. The Showcase is a recurring semiannual event designed to provide a forum that allows Duke staff, students, trainees, and faculty to highlight their work, research, or class projects, while also gaining experience in presenting their projects to a larger audience. This year’s event encompassed over 60 posters on topics that included statistics, biomedical informatics, epidemiology, machine learning, implementation science, process engineering, quality improvement, among many others.

READ MORE

Earlier this month, Duke University Executive Vice President for Health Affairs and School of Medicine Dean Mary Klotman, MD, announced the launch of Discovery AI, a new strategic research initiative designed to accelerate the application of artificial intelligence and machine learning to discovery science in biomedicine. Discovery AI is focused on deeply integrating computational and biological expertise with the goal of extending the frontiers of biological discovery across multiple scales. Scott H. Soderling, PhD, chair of the Department of Cell Biology, interim chair of the Department of Biochemistry, and David Page, PhD, chair of the Department of Biostatistics and Bioinformatics, will lead an effort that combines excellence in biology, engineering, and data science in a seamless and collaborative framework, bringing quantitative and experimental approaches into closer alignment. Duke AI Health Faculty Council member Ricardo Henao, PhD, is among numerous faculty experts who will be contributing to the initiative.

LEARN MORE

Our 2025 Duke AI Health Friday Roundup series continued to offer a weekly collection of notable news and views on AI, clinical research, basic science, and more, drawn from the scholarly literature, tech reporting, and broader media.

In December, Duke AI Health’s Community of Practice, in partnership with the Duke Pratt School of Engineering, the Duke Center for Computational and Digital Health Innovation, and Duke Health, hosted the 2025 Health Data Science Poster Showcase. The event, which was open to Duke students, trainees, staff, and faculty, also included participants in the Fall 2025 AI Health Foundations of Scientific Writing for Staff Workshop.

This fall, the AI Health Community of Practice held its third Scientific Writing Workshop for Duke staff members. This interactive hybrid mini-course, held over multiple sessions in October and November, introduced learners to the process of writing for scholarly publication and culminated in an in-person poster presentation that was part of the December Health Data Science Poster Showcase.

October of 2025 saw a successful second iteration of the annual Duke Summit on AI for Health Innovation, co-sponsored by Duke AI Health, the Duke Center for Computational and Digital Health Innovation, and the Duke Clinical Research Institute. This year’s event took place over two days and was hosted at the NC Biotechnology Center in Research Triangle Park. You can read more about the Summit in our Impact Report.

The 2025 Medical Imaging & AI Proposal Studios selected three outstanding proposal concepts from new investigators for intensive scientific feedback and design support. During November studio sessions, investigators collaborated with some of Duke’s leading data-science experts to strengthen their proposals for upcoming high-impact funding opportunities.

Earlier this year, Duke’s Collaborative to Advance Health Equity (CACHE) Initiative awarded 1-year grants for three proposals to improve care in the Duke Health system. Focused on addressing areas where current approaches to care and support could be improved, the three projects include identification and treatment of chronic kidney disease, transitions of care for patients with opioid use disorder, and improving screening for lung cancer.

A story published by the Duke Center for Autism and Brain Development highlights work by Duke AI Health Director of Data Science Ben Goldstein, PhD, and colleagues, whose AURORA study is using machine learning to sift electronic health records to better identify and understand autism in children.

This past summer saw the return of the popular Machine Learning Summer School program led by AI Health Faculty Council member Ricardo Henao, PhD, with offerings focused on generative AI.

A Duke AI Health Faculty Affiliate Monica Agrawal, PhD, was among this year’s honorees when the 2025 Whitehead Scholars were announced by the Duke University School of Medicine in May. The program recognizes and supports outstanding potential in early-career biomedical researchers.

In May of 2025, the Patient-Centered Outcomes Research Institute awarded funding to a Duke Health program designed to implement electronic monitoring of patients’ self-reported symptoms during cancer treatment. AI Health Advisory Board Member Richard Shannon, who also serves as chief medical officer and chief quality officer for Duke Health, and Duke hematologist/oncologist Thomas LeBlanc, are leading the study.

AI Health Data Science Fellowship Director Matthew Engelhard, PhD (R) is featured in a Duke story reporting on a model developed by Dr. Jonathan Posner, former AI Health Data Science Fellow Elliot Hill, and Engelhard at Duke that uses machine learning to predict mental health risks in adolescents. The project is supported by a $15 million grant from the National Institute of Mental Health.

In January of 2025, former Duke AI Health Director Michael Pencina, PhD (L) is interviewed by JAMA + AI editor Roy Perlis and discusses his recent publication on AI tracking tools for transparency and collaboration. Also in January, Duke AI Health Faculty Council member Michael Cary, PHD, RN (R) is featured in Duke Magazine.

Duke AI Health Director of Data Science Ben Goldstein, PhD, is among the authors of a new retrospective cohort study that examines relationships between patient outcomes during recovery from traumatic brain injuries and a set of social determinants of health, assessed at the level of the neighborhood environment. The article, titled “Association of neighborhood disadvantage with clinical and healthcare utilization outcomes following traumatic brain injury,” is available online ahead of print from the Journal of Clinical Neuroscience.

People are increasingly seeking healthcare information from large language models (LLMs) via interactive chatbots, yet the nature and inherent risks of these conversations remain largely unexplored. Recent research led by Monica Agrawal, PhD, a Duke AI Health faculty affiliate, releases HealthChat-11K, a curated dataset of 11K real-world chatbot conversations in which users seek healthcare information. This dataset can be used to analyze user interactions, including dangerous interactions with the potential to induce sycophancy in LLMs. The paper, titled “‘What’s Up, Doc?’: Analyzing How Users Seek Health Information In Large-Scale Conversational AI Datasets”, was presented in November as a Findings paper at the Conference on Empirical Methods in Natural Language Processing (EMNLP).

READ MORE

In a new research article titled “Using mixture cure models to address algorithmic bias in diagnostic timing: autism as a test case,” group of Duke authors, including AI Health Director of Data Science Ben Goldstein, PhD, and Data Science Fellowship Director Matthew Engelhard, PhD, examine algorithmic approaches for models used to predict autism diagnosis. The simulation study, which is published in the journal JAMIA Open, suggests that mixture cure models show promise in improving predictive modeling in autism and potentially other conditions.

READ HERE

Join the ORBIT (Observational Research Building Interdisciplinary Therapeutic Advances) Interdisciplinary Hub for a three-day Winter School in Real-World Data Analytics, January 28–30, 2026. This virtual event is open to anyone interested in real-world data analytics, regardless of affiliation with Duke University. Over three days, attendees will explore the promises and pitfalls of artificial intelligence in clinical research using real-world data, review foundational methods of causal inference and econometrics, and examine the interplay between real-world data and clinical trial design, conduct, and analysis. Speakers include AI Health Faculty Council’s Ricardo Henao and Fan Li, AI Health Faculty Affiliates Monica Agrawal and Chuan Hong, and the following faculty and experts: David Carlson, Feng Gao, Jay Lusk, Brian Mac Grory, Ryan McDevitt, Emily O’Brien, Dylan Thibault, Laine Thomas, Chengxin Yang, and Anqi Zhao.

LEARN MORE

Inventions and emerging technologies developed at Duke were on display at the seventh annual Invented at Duke celebration, held at Duke’s Penn Pavilion on November 11. The showcase, which is hosted and sponsored by Duke’s Office of Translation and Commercialization, puts a spotlight on the university’s innovation and entrepreneurship ecosystem and offers networking opportunities for numerous featured resources – Duke AI Health among them!

READ MORE

A research article published in the journal Neonatology and featuring a group of authors from Duke’s Department of Pediatrics and AI Health’s Matthew Engelhard, PhD, and Ricardo Henao, PhD, explores the use of machine learning models to improve risk predictions for retinopathy of prematurity. The article, titled “Machine Learning Risk Prediction for Treated Retinopathy of Prematurity in Infants,” appears online ahead of print.

READ HERE

Following this October’s Duke Summit on AI for Health Innovation, co-hosted by Duke AI Health, Duke’s Center for Computational and Digital Health Innovation, and the Duke Clinical Research Institute, Duke AI Health Research Scientist Whitney Welsh, PhD, has compiled an impact report distilling some key information from the two-day convening, including breakdowns of conference attendees, insights from attendees, and complete lists of speakers, panelists, partners, sponsors, and discussion group representatives.

READ MORE

The 2025 Medical Imaging & AI Proposal Studios selected three outstanding proposal concepts from new investigators for intensive scientific feedback and design support. During November studio sessions, investigators collaborated with some of Duke’s leading data-science experts to strengthen their proposals for upcoming high-impact funding opportunities. This effort reflects our broader commitment to accelerating innovative research at Duke through expert mentorship, interdisciplinary collaboration, and strategic proposal development.

LEARN MORE

The Collaborative to Advance Clinical Health Equity (CACHE) at Duke Health is soliciting applications for innovative projects that leverage CACHE’s infrastructure and capabilities to identify, address, and eliminate disparities in healthcare. This RFA seeks interdisciplinary projects that utilize data science, comparative effectiveness research, predictive modeling, quality improvement, implementation science, social epidemiology, and community engagement to identify and mitigate healthcare disparities. Selected projects will receive substantial analytics, informatics, community engagement, quality improvement framework didactic training, QI engineering support, and mentorship from the CACHE team. In addition, CACHE will work with health system leaders to provide project management, informatics, and statistical support.

LEARN MORE

Earlier this spring, Duke’s Collaborative to Advance Health Equity (CACHE) Initiative awarded 1-year grants for three proposals to improve care in the Duke Health system. Focused on addressing areas where current approaches to care and support could be improved, the three projects include identification and treatment of chronic kidney disease (CKD), transitions of care for patients with opioid use disorder (OUD), and improving screening for lung cancer, all of which pose substantial challenges among patients served by Duke Health. “We wanted to focus on work that brings together elements of care that have been shown to be effective but haven’t yet been widely integrated into clinical workflows or scaled,” notes CACHE Director Michael Pignone, MD, MPH.

READ MORE

In a research article published this month in the journal BMC Medical Imaging, Duke AI Health Faculty Council Member Ricardo Henao joins first author Conor Artman to describe a new approach for using deep learning models to detect clinically significant data in echocardiograms. Unlike commonly used but data-intensive segmentation models, the authors propose a “Scaled Gumbel Softmax” deep learning model that offers improvements over existing models while requiring fewer resources.

READ HERE

In an article by Mariah Drexler and Elaine Xiao that originally appeared in the Duke Chronicle, former AI Health Director Michael Pencina, PhD, and AI Health Faculty Affiliate Chuan Hong, PhD, are interviewed about their work in creating SCRIBE, a framework for evaluating AI applications that generate real-time notes during patient encounters in a hospital or other clinical setting.

READ MORE

We are excited to announce the launch of the Duke AI Health Industry Studios program. Powered by AI experts in Duke’s Department of Biostatistics & Bioinformatics, we offer half- or full-day design workshops tailored to an organization’s strategy, problem-solving, and solutions development. Our Duke faculty conduct research on the frontiers of health AI methods and real-world applications and can help partners develop solutions to their toughest challenges, no matter where they are in AI implementation, from data collection and management to model development, or from measuring model performance to ongoing monitoring.

LEARN MORE

Mark your calendars! Duke AI Health, in partnership with Duke Health and the Pratt School of Engineering, will be hosting the Health Data Science Poster Showcase on Friday, December 12, from 11:30 AM to 1:30 PM at Duke’s Fitzpatrick Center for Interdisciplinary Engineering, Medicine and Applied Sciences (FCIEMAS) Atrium, Ground Level. No registration is needed for this event, which is free and open to the public. We encourage everyone to come by and catch a glimpse of some of the innovative ideas in health data science percolating at Duke!

Join Duke AI Health for a virtual seminar, “The Protected Research Compute Cluster: Introduction for School of Medicine Researchers,” on Tuesday, November 11, 2025, from 12–1 p.m. ET. This session will introduce the Protected Research Compute Cluster (PRCC)—a new, secure high-performance computing environment that will replace PACE with expanded capacity, improved reliability, and scalable technology for sensitive data research. Presenters include Ricardo Henao, PhD, Aby Veiga, Jay Stotler, Danny Williford and John Bradley, and the session will be hosted by Shelley Rusincovitch, MMCi.

REGISTER

This past October, Duke AI Health, in partnership with the Center for Computational and Digital Health Innovation at Duke University and the Duke Clinical Research Institute (DCRI) hosted the 2nd annual Duke Summit on AI for Health Innovation at the North Carolina Biotechnology Center in Research Triangle Park. The conference, which took place October 8-9, brought together experts representing a wide range of fields including healthcare, engineering, computer science, and the biomedical sciences to explore how AI can drive health innovation. Through a series of presentations, lightning talks, panel discussions, and structured “breakout” tables, participants were invited to:

• Engage in conversations about the future agenda of AI-driven health innovation;
• Network with leaders from academia, industry, and healthcare;
• Understand how to work with healthcare & learn about AI limitations and opportunities; and
• Learn about the landscape of AI development in healthcare.

READ MORE

Duke AI Health Faculty Affiliate and Data Science Fellowship Director Matthew Engelhard is among the Duke faculty interviewed for a Duke Chronicle story by Lucas Lin and Ananya Pinnamaneni on the use of artificial intelligence in university learning environments. The article, which describes a pilot partnership between Duke and OpenAI to study the classroom use of ChatGPT, was later picked up by Associated Press and WRAL.

READ MORE

Michael Pencina, PhD, has stepped down as vice dean for data science at Duke University School of Medicine and chief data scientist for Duke Health, effective September 30, 2025. Dr. Pencina accepted an exciting new opportunity as chief AI scientist with UnitedHealth Group. We will miss his leadership and celebrate the extraordinary foundation he has built for our school’s future in data science and artificial intelligence.

Over the past decade, Dr. Pencina has been instrumental in shaping Duke’s data science strategy for research, clinical care, and education. His leadership positioned Duke as a national leader in the responsible and rigorous application of AI in health care. Through initiatives like Duke AI Health and the Algorithm-Based Clinical Decision Support Oversight program, he has embedded governance, transparency, and trust into our approach to algorithmic innovation. As co-founder of the Coalition for Health AI (CHAI), he helped unite academic, industry, and federal stakeholders to promote trustworthy AI across the health care ecosystem.

A team at Duke University School of Medicine has received a $15 million grant from the National Institute of Mental Health to improve and expand an artificial intelligence (AI) tool that helps catch early signs of mental health problems in teenagers and adolescents. The AI model, called the Duke Predictive Model of Adolescent Mental Health (Duke-PMA), analyzes data on behavior, emotions, and brain function to identify kids at high risk for mental illness even before symptoms appear. It looks at a range of easy to measure factors, like sleep patterns and family stress, and has already shown it can predict worsening mental health up to a year in advance with 84% accuracy in kids ages 10 to 15.

READ MORE

Original story by Susan Gallagher, Duke University School of Medicine

At his Techy Surgeon Substack page, Duke orthopedic surgeon and Duke Margolis Institute core faculty member Christian Péan, MD, describes his upcoming presentation at the Duke AI Summit on AI for Health Innovation, happening October 8-9, 2025, at the NC Biotechnology Center:

“I’ll be speaking about how AI-driven workflows close care gaps and improve patient experience, with a focus on agentic systems that escalate intelligently, create auditable loop closure, and support value-based care without adding burden.” 

LEARN MORE

Earlier this month, Duke AI Health Director Michael Pencina and AI Health Communications Director Jonathan McCall were among a number of Duke faculty and staff participating in a conference on the impact of AI technologies on work, education, personal development and more. The conference, titled “The Human Edge: Our Future with Artificial Intelligence” was co-presented by Elon University and RTI and took place on RTI campus in Research Triangle Park. Elon University’s Lee Rainie also debuted findings from a recent report surveying attitudes and perceptions related to AI.
WATCH

We’re excited to announce the launch of the Protected Research Compute Cluster (PRCC), a high-performance, secure environment for research involving PHI and other sensitive data. PRCC offers customizable workspaces with familiar tools, optimized GPU and storage capacity, and simplified access to Duke Health’s data resources. Replacing PACE, PRCC expands capacity and reliability while enabling global collaboration, all within a NIST 800-53 compliant framework. Together with the Research Computing Cluster (RCC), PRCC forms part of the Duke University School of Medicine Research Enclave, giving researchers flexible, scalable options for secure data analysis.

LEARN MORE

Andrew Olson, Associate Director of Policy Strategy and Solutions for Health Data Science at Duke AI Health, will present for the Office of Physician-Scientist Development’s Research Careers Ahead Virtual Seminar Series in a virtual seminar via Zoom on Wednesday, October 22, 2025, from 4:00–5:00pm ET. His talk, “Clinical Risk Prediction Modeling with Machine Learning and AI,” will explore how these methods are being applied to improve health outcomes.

LEARN MORE

Patients are increasingly using generative AI to answer health questions, through tools like chatbots or AI-powered search results. Recent research led by Monica Agrawal, PhD, AI Health Faculty Affiliate and Assistant Professor of Biostatistics & Bioinformatics, characterizes the potential failure modes of this phenomenon, analyzes how LLM-generated responses can mislead patients even without hallucinations, and offers recommendations for building safer systems. The paper, “Retrieval-augmented systems can be dangerous medical communicators,” was presented in July at the International Conference on Machine Learning (ICML).

LEARN MORE

We invite you to a virtual seminar on Tuesday, September 23, 2025, from 12:00–1:00pm ET for a virtual seminar via Zoom, open to all internal and external participants. Xiaoyue Ni, PhD, Assistant Professor of Mechanical Engineering & Materials Science and Biostatistics & Bioinformatics at Duke University, will present on a soft, skin-mounted mechano-acoustic (MA) sensing platform that records body sounds and kinematics with high fidelity. This technology leverages epidermal electronics to capture a high-dimensional array of mechanical and acoustic signatures, enabling comfortable, accurate, and comprehensive decoding of physiological states, behavioral patterns, functional performance, and cognitive or intentional states in real time.

LEARN MORE

We invite you to attend a virtual seminar taking place on Thursday, October 2 from 12:00 noon to 1:00 PM Eastern, where Duke AI Health Faculty Affiliate Chuan Hong, PhD, will share findings from a large-scale evaluation of the American Heart Association’s PREVENT model for predicting 10-year cardiovascular risk. She will discuss the model’s performance across diverse populations and the impact of social determinants of health, highlighting both its strengths and key disparities. Her talk will consider the implications for applying PREVENT in real-world clinical settings. The seminar is free and open to the public.

LEARN MORE

Clinical notes often contain important descriptive findings not captured in structured EHR fields, making them valuable for early autism prediction. However, identifying autism-related insights is difficult due to their sparsity within the large volume of notes for a typical child. Duke researchers, including Computational Biology & Bioinformatics student Fengnan Li, AI Health Data Science Fellow Elliot Hill, and Duke AI Health Data Science Fellowship Director Matthew Engelhard, PhD have developed a new natural language processing method, IRIS (Interpretable Retrieval-Augmented Classification for long Interspersed Document Sequences), to address this challenge. Their work was recently published at the 2025 Annual Meeting of the Association for Computational Linguistics.

LEARN MORE

AI Health Director of Data Science Ben Goldstein, PhD, and AI Health Faculty Affiliate Matt Engelhard, MD, PhD, will be presenting papers at the upcoming Machine Learning for Healthcare (MLHC) conference at the Mayo Clinic in Rochester, MN. The first paper, Borrowing from the Future: Enhancing Early Risk Assessment through Contrastive Learning (Sun, Engelhard, Goldstein), explores improved early risk prediction using contrastive learning methods. The second, Balancing Interpretability and Flexibility in Modeling Diagnostic Trajectories with an Embedded Neural Hawkes Process Model (Zhao, Engelhard), investigates modeling approaches that support both transparency and complexity in clinical data. Duke AI Faculty Council Member Ricardo Henao, PhD, will also be presenting a poster with colleague Mohd Ashhad titled Generating Accurate Synthetic Survival Data by Conditioning on Outcomes. LEARN MORE

Join us on Wednesday, September 17, 2025, from 4:00–5:00pm ET for a virtual seminar via Zoom, open to all internal and external participants. Monica Agrawal, PhD, Assistant Professor of Biostatistics & Bioinformatics at Duke University, will explore how natural language processing and large language models are transforming clinical text analysis to advance research, streamline physician workflows, and improve patient access to information. She will discuss scalable information extraction, smarter electronic health records, evaluation challenges for generative AI in medicine, and patient use of language models for health information.

REGISTER

As Duke AI Health and the Pratt School of Engineering prepare to hold the second Duke AI for Health Innovation Summit in October (see Events section below), get inspired by the engagement and discussion of last year’s summit conference proceedings! The white paper available at the link below (PDF) captures key presentations, panel discussions, and informal “fireside chats.” READ MORE

Registration Open for Duke Summit on AI for Health Innovation: October 8-9, 2025!

“’The question is, again, what are the consequences of it?’ The answer, to him, rests in the stakes of making an error – and with healthcare, those stakes are serious.” The Verge’s Hayden Field interviews AI Health Director Michael Pencina, PhD, for an article that probes the implications of recent reports suggesting that a specialized medical large language model chatbot – Google’s Med-Gemini – may have hallucinated a nonexistent anatomical feature. The lapse, which made its way unrecognized into a preprint paper posted by Google, has raised concerns among AI researchers.

READ MORE

Duke AI Health partners with investigators across the university on a wide range of impactful research. We’re excited to launch our new Project Profile feature, designed to spotlight this important work and showcase interdisciplinary innovation. Explore the latest profiles and learn more about these collaborations at the link below. READ MORE

A group of authors from Duke, including AI Health Data Science Fellow Elliot Hill and Data Science Fellowship Director Matthew Engelhard, published a research article that examined the propensity for a supervised machine learning approach known as binary classification to yield biased results when predicting long-horizon diagnoses. The paper, titled “Limitations of Binary Classification for Long-Horizon Diagnosis Prediction and Advantages of a Discrete-Time Time-to-Event Approach: Empirical Analysis,” was published in March in the journal JMIR AI.
READ HERE

Duke Office of Information Technology Media Architect and Senior Producer Stephen Toback explores the capabilities and limitations of AI chatbots like ChatGPT in healthcare and academia. The session demonstrates real-world use cases ranging from clinical documentation to administrative support. Toback also addresses common pitfalls, such as hallucinations and privacy concerns, providing guidance on safe implementation. It’s a practical guide for institutions considering integrating generative AI tools. 

WATCH